Acquisition of Dynamic Function in Human Stem Cell-Derived β Cells

Highlights

• •Development of differentiation protocol to β-like cells with enhanced function
• •TGF-β signaling promotes acquisition of dynamic function in maturing β-like cells
• •Transplanted cells rapidly restore glucose tolerance in mice

Summary

Recent advances in human pluripotent stem cell (hPSC) differentiation protocols have generated insulin-producing cells resembling pancreatic β cells. While these stem cell-derived β (SC-β) cells are capable of undergoing glucose-stimulated insulin secretion (GSIS), insulin secretion per cell remains low compared with islets and cells lack dynamic insulin release. Herein, we report a differentiation strategy focused on modulating transforming growth factor β (TGF-β) signaling, controlling cellular cluster size, and using an enriched serum-free media to generate SC-β cells that express β cell markers and undergo GSIS with first- and second-phase dynamic insulin secretion. Transplantation of these cells into mice greatly improves glucose tolerance. These results reveal that specific time frames for inhibiting and permitting TGF-β signaling are required during SC-β cell differentiation to achieve dynamic function. The capacity of these cells to undergo GSIS with dynamic insulin release makes them a promising cell source for diabetes cellular therapy.

Graphical Abstract

Keywords

• human embryonic stem cells human induced pluripotent stem cells diabetes differentiation glucose-stimulated insulin secretion
• transplantation cell therapy β cells pancreas islets

Introduction

Nostro et al., 2015, Rezania et al., 2012), the latter of which can be transplanted into rodents and spontaneously differentiated into β-like cells after several months (Bruin et al., 2015, Kroon et al., 2008

Millman et al., 2016

Rezania et al., 2012

Pagliuca et al., 2014

Rezania et al., 2014

Here we report a six-stage differentiation strategy that generates almost pure populations of endocrine cells containing β-like cells that secrete high levels of insulin and express β cell markers. This is achieved by modulating Alk5i exposure to inhibit and permit TGF-β signaling during key stages in combination with cellular cluster resizing and enriched serum-free media (ESFM) culture. These cells are glucose responsive, exhibiting first- and second-phase insulin release, and respond to multiple secretagogues. Transplanted cells greatly improve glucose tolerance in mice. We identify that inhibiting TGF-β signaling during stage 6 greatly reduces the function of these differentiated cells while treatment with Alk5i during stage 5 is necessary for a robust β-like cell phenotype.

Results

 Differentiation to Glucose-Responsive SC-β Cells In Vitro

We set out to develop an improved differentiation protocol starting from the approach we described in 

Pagliuca et al. (2014) using the HUES8 cell line. We included Y27632 during stages 3 to 4 and activin A during stage 4 as we reported previously (Millman et al., 2016) to help maintain cluster integrity and shortened stage 3 from 2 days to only 1 day to enhance progenitors (Nostro et al., 2015). We also developed an ESFM for stage 6 to replace the serum-containing media used previously to have a serum-free protocol. During our protocol pilot studies, we observed that both resizing clusters and removal of Alk5i and T3 increased insulin secretion while maintaining the C-peptide+ population (Figures S1A and S1B).Combining these modifications resulted in our six-stage differentiation protocol outlined in Figure 1A. Stage 6 cells are grown as clusters in suspension culture (Figure 1B) that averaged 172 ± 34 μm (mean ± SD; n = 353 individual clusters) in diameter, less than half the diameter of the clusters before resizing, which was 364 ± 55 μm (n = 155 individual clusters). Stage 6 clusters stained red for the zinc-chelating dye dithizone, which stains β cells. Immunostaining of sectioned clusters revealed most cells to be C-peptide+, a protein also produced by the INS gene, in addition to PDX1+ and NKX6-1+, β cell markers (Figure 1C). A subset of cells stained positive for GCG or were polyhormonal, staining positive for both C-peptide and GCG. These polyhormonal cells are known to not to resemble adult β cells and are not functional (D'Amour et al., 2006, Hrvatin et al., 2014).

We tested function of stage 6 cells generated with our differentiation protocol using both static (Figures 1D, 1E, and S1C) and dynamic GSIS assays (Figures 1F and S1D) and found that not only do the cells secrete insulin but also increase insulin release when moved from low to high glucose. With static GSIS, while there was some variability, stage 6 cells increased insulin secretion on average by a factor of 3.0 ± 0.1 when moved from 2 to 20 mM glucose. This is an improvement compared with cells generated with the protocol described in 

Pagliuca et al. (2014)

 (1.4 ± 0.1), referred to here as the Pagliuca protocol, but less than human islets (3.2 ± 0.1) on average (Figure 1D). Stage 6 cells from this study did not increase insulin secretion in response to 5.6 mM glucose but did increase secretion in response to higher concentrations (11.1 and 20 mM), indicating that the cells are only stimulated at higher glucose threshold (Figure 1E). In terms of insulin secretion per cell, stage 6 cells secreted on average 5.3 ± 0.5 μIU/10³ cells at 20 mM glucose, 9.2 ± 1.1 times more than cells generated with the Pagliuca protocol and 2.3 ± 0.3 times less than human islets, on average (Figure 1D). It is important to note that our insulin values with the Pagliuca protocol are within range of the 2014 report but were lower on average, with differentiated HUES8 reported to secrete 0.2–2.6 μIU/10³ cells (average 1.4) and increase secretion by 0.4–4.1 (average 1.7) to high glucose.With dynamic GSIS, stage 6 cells displayed a rapid first-phase insulin release within 3–5 min of high glucose exposure, increasing insulin secretion by a factor of 7.6 ± 1.3 to 159 ± 21 μIU/μg DNA, higher than stage 6 cells generated from the Pagliuca protocol (1.7 ± 0.2× increase to 11 ± 1 μIU/μg DNA) but lower than human islets (15.0 ± 2.4× increase to 245 ± 26 μIU/μg DNA) (Figure 1F). Second-phase insulin secretion was observed with continued high glucose exposure, with cells maintaining 2.1 ± 0.3 higher insulin secretion than the initial low glucose, a higher increase than with the Pagliuca protocol (0.9 ± 0.1) but lower than human islets (6.7 ± 0.8) (Figure 1F). When the cells were returned to low glucose, insulin secretion from stage 6 cells returned to a reduced rate. Elevating insulin secretion and displaying first- and second-phase insulin release to a high glucose challenge are key features of β cell behavior. Overall, stage 6 cells generated with this differentiation strategy produces cells with clear first- and second-phase insulin secretion, which was not demonstrated by Pagliuca et al. (2014)and Rezania et al. (2014)
 and not seen with stage 6 cells produced with the Pagliuca protocol. However, when compared with human islets containing β cells, these stage 6 cells still have lower insulin secretion per cell at high glucose, lower glucose stimulation on average, and slightly slower first-phase insulin release.To further characterize stage 6 cells generated with our differentiation protocol, we immunostained cells with a panel of pancreatic islet markers (Figures 2A–2C and S2). The vast majority of cells expressed CHGA (96% ± 1%), a pan-endocrine marker, and most cells expressed C-peptide (73% ± 3%) (Figure 2). These fractions are higher than in stage 6 cells generated with the Pagliuca protocol (Figure S2) and reported in Pagliuca et al. (2014). Many C-peptide+ cells from both protocols expressed other markers found in β cells and expression of the other pancreatic hormones was observed (Figures 2 and S2). The majority of C-peptide+ cells expressed NKX6-1 (Figure 2) and were monohormonal, which we presumed to be the SC-β cell population as done previously (Pagliuca et al., 2014). The fraction of C-peptide+ cells not expressing another hormone was increased compared with stage 6 cells generated with the Pagliuca protocol and reported in Pagliuca et al. (2014)
, while the fraction of these cells expressing another hormone was comparable (Figures 2 and S2). These data show that stage 6 cells generated with our differentiation strategy are predominantly pancreatic endocrine with the majority expressing C-peptide.

We measured expression of several genes compared with stage 6 cells generated with the Pagliuca protocol and human islets (Figures 2D and S3). Many islet and β cell genes were increased compared with the Pagliuca protocol, including INS, CHGA, NKX2-2, PDX1, NKX6-1, MAFB, GCK, and GLUT1. Interestingly, LDHA and SLC16A1, disallowed β cell genes, had reduced expression in our stage 6 cells compared with both the Pagliuca protocol and human islets (LDHA) and the Pagliuca protocol (SLC16A1). Our stage 6 cells had increased expression of CHGA, NKX6-1, MAFB, GCK, and GLUT1 compared with human islets. However, INS, GCG, SST, and particularly MAFA and UCN3 had reduced expression compared with stage 6 cells. However, several recent reports have provided evidence that question the utility of MAFA and UCN3 in evaluating human SC-β cell maturation. MAFA expression is low in juvenile human β cells (Arda et al., 2016). MAFB is expressed in human but not mouse β cells (Arda et al., 2016, Tritschler et al., 2017, Xin et al., 2016). UCN3 expression is much higher in mouse than human β cells (

Xin et al., 2016) and is also expressed by human α cells (Baron et al., 2016, Tritschler et al., 2017). These data show that our stage 6 cells have improved gene expression for many markers compared with the Pagliuca protocol and, while the expression of several β cell markers are equal to or great than human islets, other markers remain low.

 Transplantation of SC-β Cells into Glucose-Intolerant Mice

To evaluate the functional potential of stage 6 cells in vivo, we first transplanted cells under the renal capsule of non-diabetic mice and evaluated the ability of the graft to respond to a glucose challenge (Figure 3A). We observed that, even after extended time post-transplantation (6 months), the grafts responded to a glucose injection by increasing human insulin by a factor of 1.9 ± 0.5. Excision and immunostaining of the transplanted kidneys revealed C-peptide+ cells that tended to be clustered together in addition to other pancreatic endocrine and exocrine markers (Figures 3B and S4A). To more rigorously evaluate stage 6 cells in vivo, we transplanted a separate mouse cohort that had been chemically induced to be diabetic with streptozotocin (STZ) and evaluated function at early (10 and 16 day) and late (10 week) time points. After only 10 days post-transplantation, STZ-treated mice receiving stage 6 cells had greatly improved glucose tolerance compared with STZ-treated sham mice and had similar glucose clearance as the non-STZ-treated mice (Figures 3C and 3D). Measurements of human insulin 16 days after transplantation revealed high insulin concentrations that increased by a factor of 2.3 ± 0.6 with a glucose injection to 16.6 ± 3.1 μIU/mL (Figure 3E). These values are greater than what was reported in Pagliuca et al. (2014)
 under similar conditions, which had an insulin increase of 1.4 ± 0.3 and concentration of 3.8 ± 0.8 μIU/mL. Observing our cohort 10 weeks after transplantation revealed similar results as the 10- and 16-day data, with transplanted mice having greatly improved glucose tolerance (Figures 3F and 3G) and glucose-responsive insulin secretion (Figure 3H). Mice not receiving STZ had similar glucose tolerance as mice receiving a therapeutic dose of human islets (Pagliuca et al., 2014). Mice that did not receive stage 6 cells had undetectable human insulin and mice that received STZ had drastically reduced mouse C-peptide compared with non-STZ-treated mice (Figures S4B and S4C). Grafts from these STZ-treated mice contained cells that expressed β cell markers in addition to other endocrine and exocrine markers (Figure S4D). Overall these data demonstrate that stage 6 cells generated with our protocol are functional both at early and late time points in vivo, greatly improving glucose tolerance to equal that of non-STZ-treated mice.

 Characterization of SC-β Cell Dynamic Function

Since the differentiation protocol produces cells that are capable of dynamic insulin secretion, we studied this phenotype in more detail. We performed dynamic GSIS on cells as they progressed through stage 6 (Figure 4A). We observed that robust dynamic function was transient, with cells at 5 days secreting low amounts of insulin and exhibiting weak first- and second-phase response, with later time points (9–26 days) secreting higher amounts of insulin with a clear first- and second-phase response. During this time, the fraction of C-peptide+ cells decreased slightly (Figure S5A). By 35 days, insulin secretion at low glucose had risen such that first and second phases were difficult to clearly identify. These data show that SC-β cells require 9 days in stage 6 to acquire dynamic function, this function persists for weeks, but after extended in vitro culture glucose responsiveness is lost. Similarly, cadaveric human islets are known to have a limited functional lifetime in vitro, but the cause of this is not clear. These data further suggest an optimal time frame for these cells to be used in transplantation and drug-screening studies.

To further characterize dynamic insulin secretion, we performed perifusion experiments to assay whether SC-β cells could respond to sequential challenges with several known secretagogues (Figure 4B). After an initial high glucose challenge, SC-β cells were able to respond to a second high glucose-only challenge, albeit less strongly than the first challenge, and extending the first glucose challenge to 1 hr in a separate experiment did not reduce insulin secretion (Figure 4C). Addition of other secretagogues during the second challenge further increased insulin secretion (Figure 4B). Membrane depolarizers KCl and L-arginine had the largest increases. Tolbutamide (blocks potassium channel), 3-isobutyl-1-methylxanthine (IBMX) (raises cytosolic cAMP), and exendin-4 (agonist of GLP-1 receptor) also increased insulin secretion over high glucose alone. Not only was insulin secretion increased but it rose faster than with high glucose alone. However, we noted that the response of stage 6 cells to KCl challenge was stronger than in human islets (Figure S5B), an observation made by others comparing β-like cells to human islets (

Rezania et al., 2014

), possibly indicative of continued immature or juvenile β cell phenotype. Taken together, these data show that SC-β cells can respond to several secretagogues that have diverse modes of action and have potential application in drug screening.

 Role of TGF-β Signaling in SC-β Cell Differentiation and Maturation

After having evaluated SC-β cells generated with our protocol, we investigated the protocol changes that were made to gain insights into SC-β cell differentiation and maturation. We found that, while inclusion of Alk5i during stage 6 resulted in relatively weak but statistically significant GSIS in a static assay, similar to data from the Pagliuca protocol (Figure 1D), omission of Alk5i drastically increased insulin secretion and glucose stimulation (Figures 5A and S6A). Insulin content also increased with removal of Alk5i during stage 6 (Figure 5B), but the proinsulin/insulin ratio remained similar (Figure 5C), suggesting that the increased insulin content is not due to hormone processing. Furthermore, the fraction of cells expressing pancreatic endocrine markers, including C-peptide, remained similar between DMSO- and Alk5i-treated cells (Figures 5D, 5E, and S6B). Gene expression was similar overall with and without Alk5i treatment, with cluster resizing typically having a larger effect (Figure S6C). Cells treated with Alk5i during stage 6 also had dramatically reduced insulin secretion with the dynamic GSIS assay, displaying weak to no first- and second-phase response (Figure 5F), similar to cells generated with the Pagliuca protocol (Figure 1F). These data show that Alk5i treatment during stage 6 inhibits functional maturation of SC-β cells.

Our studies with Alk5i during stage 6 suggested that permitting TGF-β signaling was necessary for robust functional maturation of SC-β cells, as inhibition of TGFBR1 is the canonical function of Alk5i. To test this hypothesis, we first used western blot analysis to validate that TGF-β signaling was occurring in our stage 6 cells via SMAD phosphorylation (Figure 6A). Alk5i treatment diminished phosphorylated SMAD, confirming that TGF-β signaling was indeed occurring and inhibited by Alk5i. SMAD phosphorylation was observed in stage 6 clusters regardless of whether they were resized, consistent with observations that Alk5i treatment reduced GSIS regardless of resizing (Figure S7). Next, we generated two lentiviruses carrying short hairpin RNA (shRNA) designed to knock down TGFBR1 (TGFBR1 no. 1 and no. 2). These viruses were capable of reducing TGFBR1 transcript compared with control virus targeting GFP in stage 6 cells (Figure 6B) and reduced SMAD phosphorylation (Figures 6C and S7C), albeit to much lesser extent than Alk5i treatment (Figure 6A). Similar to Alk5i treatment (Figures 5A and 5F), stage 6 cells transduced with shRNA against TGFBR1 had reduced insulin secretion and reduced positive glucose responsiveness in the static GSIS assay (Figure 6C) and blunted glucose response in the dynamic GSIS assay (Figure 6D). These data show that permitting TGF-β signaling during stage 6 is important for SC-β cell functional maturation, which is inhibited by treatment with Alk5i.

Finally, we studied the role of Alk5i during stage 5 of differentiation to evaluate its effects on differentiation toward pancreatic endocrine cells, as it had been used previously for endocrine induction (Millman et al., 2016, Pagliuca et al., 2014
, Rezania et al., 2014, Russ et al., 2015, Zhu et al., 2016). These experiments were performed as outlined in Figure 1A in the presence or absence of Alk5i. We observed that the fraction of cells differentiated to endocrine cells (CHGA+) was unchanged but the fraction of cells differentiated to a C-peptide+ phenotype was decreased by omitting Alk5i (Figures 7A–7C). Similarly, the fraction of cells co-expressing C-peptide and NKX6-1, an important transcription factor for specifying β cells (Rezania et al., 2013, Rieck et al., 2012), was decreased by omitting Alk5i. INS and GCG gene expression decreased with Alk5i omission, but surprisingly SST expression was slightly increased (Figure 7D). Expression of NKX6-1 and PDX1 were reduced without Alk5i (Figure 7E), while expression of several pancreatic endocrine markers were either unchanged or only slightly changed (Figure 7F). To further test the importance of Alk5i during stage 5, cells treated with or without Alk5i during stage 5 were further cultured for 7 days in stage 6 without Alk5i nor cluster resizing, and insulin secretion was substantially higher in cells treated with Alk5i during stage 5 (Figure 7G). Taken together, these data show that Alk5i treatment during stage 5 positively influences specification to β-like cell fate, not necessary to specify endocrine cells, and is necessary for high insulin secretion of resulting SC-β cells. In addition, these observations illustrate the importance of stage-specific treatment of the TGF-β signaling-inhibitor Alk5i to both generate and functionally mature SC-β cells.

Discussion

Here we demonstrate that enhanced functional maturation of SC-β cells is achieved with our six-stage differentiation strategy. These cells secrete a large amount of insulin and are glucose responsive, displaying both first- and second-phase insulin release. This differentiation procedure generates almost pure endocrine cell populations without selection or sorting, and most cells express C-peptide and other β cell markers. Upon transplantation into STZ-treated mice, glucose tolerance is rapidly restored and function persists for months. These SC-β cells respond to multiple secretagogues in a perifusion assay. We found modulating TGF-β signaling to be crucial for success, with inhibition during stage 5 increasing SC-β cell differentiation but inhibition during stage 6 reducing function and insulin content. Permitting TGF-β signaling during stage 6 was necessary for robust dynamic function.

Kayton et al., 2015, Lyon et al., 2016, Pagliuca et al., 2014). We do note that islets in our study were typically more functional than in other studies (Ghazizadeh et al., 2017, Pagliuca et al., 2014, Rezania et al., 2014, Russ et al., 2015, Sui et al., 2018), which we believe is important to rigorously benchmark SC-β cells. In addition, some islet genes remain underexpressed in our cells. Furthermore, while the data we generated with the Pagliuca protocol were within the range of data presented in the 2014 report, we acknowledge that the static GSIS values were lower on average, likely due in part to technical differences in how the assays were performed. Another potential contributor is batch-to-batch variability, as stated in the 2014 report, which could be caused by the use of different lots of serum during stage 6 and was eliminated in our protocol. Even with these difficulties and insights, we acknowledge that even further maturation of SC-β cells is possible, building on this report and the original 2014 breakthroughs.This study provides insights into the role of TGF-β signaling in functional maturation. Prior reports are unclear on this topic, with some showing TGF-β inhibition to benefit (Lin et al., 2009) and others to harm (Totsuka et al., 1989) secretion. Inhibition has been shown to promote replication (Dhawan et al., 2016), protect against stress-induced loss of phenotype (Blum et al., 2014, Millman et al., 2016), and reduce apoptosis in a GLIS3 knockout model (Amin et al., 2018). Interestingly, we also observed that removal of Alk5i during stage 5 does not affect the overall percentage of cells expressing CHGA, but influences the expression of INS, GCG, and SST, suggesting a role in TGF-β signaling in endocrine subtype specification. It is important to note that we did not identify the downstream effectors of TGF-β signaling responsible for the reported phenotypes, and further study is warranted.

Experimental Procedures

Author Contributions

L.V.C., J.S., and J.R.M. conceived of the experimental design. All authors contributed to the in vitro experiments. L.V.C., K.G.M., and J.R.M. performed all in vivo experiments. L.V.C. and J.R.M. wrote the manuscript. All authors edited and reviewed the manuscript.